Abstract
Accurate visual localization is essential for our ability to interact with the world. Previous work has shown that the perceived location of an object is influenced by its surrounding context (e.g., frames of reference, landmarks, motion), but less is known about which image statistics influence localization errors within natural scenes. We measured the influence of local image statistics (luminance, edges, object boundaries, and saliency) on perceptual reports of location. On each trial, 10 observers reported the location of a brief (50 ms) Gaussian target (σ=0.85º) superimposed on a 48º by 37º photograph of a natural scene at one of three eccentricities (5º, 7.5º, 10º) and one of 24 angular locations. Observers reported the target’s perceived location by adjusting the position of a cursor. For each statistic, we calculated the difference between the image value at the physical center of the Gaussian target and the value at its reported center, and averaged the resulting difference scores across 720 trials. To isolate image-specific effects, these difference scores were compared to a randomly-permuted null distribution, that was calculated by shuffling the mapping between the response coordinates and different images across all trials. The observed difference scores indicated that responses were significantly biased toward darker regions, luminance edges, object boundaries, and areas of high saliency (p-values <. 001; αB = .006), with low shared variance between these measures (R2 < .02). In a second experiment, 12 observers made reflexive saccades to the same targets. The results showed that the same image statistics were associated with observers’ saccade errors, despite large differences in reaction time for the two experiments (987 vs. 239 ms). Together, these results indicate that local spatial statistics influence localization in natural images, and that these biases are independent of response modality.